File Coverage

blib/lib/CGI/Buffer.pm

Criterion	Covered	Total	%
statement	60	560	10.7
branch	14	372	3.7
condition	4	219	1.8
subroutine	15	24	62.5
pod	4	4	100.0
total	97	1179	8.2

package CGI::Buffer;

use 5.14.0;	# For IO::Compress::Brotli
use strict;
use warnings;

use Digest::MD5;
use IO::String;
use CGI::Info;
use Carp;
use HTTP::Date;
use Text::Diff;	# For debugging

=head1 NAME

CGI::Buffer - Verify, Cache and Optimise CGI Output

=head1 VERSION

Version 0.84

=cut

our $VERSION = '0.84';

=head1 SYNOPSIS

CGI::Buffer verifies the HTML that you produce by passing it through
C<HTML::Lint>.

CGI::Buffer optimises CGI programs by reducing, filtering and compressing
output to speed up the transmission and by nearly seamlessly making use of
client and server caches.

To make use of client caches, that is to say to reduce needless calls
to your server asking for the same data, all you need to do is to
include the package, and it does the rest.

    use CGI::Buffer;
    # ...

To also make use of server caches, that is to say to save regenerating
output when different clients ask you for the same data, you will need
to create a cache.
But that's simple:

    use CHI;
    use CGI::Buffer;

    # Put this at the top before you output anything
    CGI::Buffer::init(
	cache => CHI->new(driver => 'File')
    );
    if(CGI::Buffer::is_cached()) {
	# Nothing has changed - use the version in the cache
	exit;
    }

    # ...

To temporarily prevent the use of server-side caches, for example whilst
debugging before publishing a code change, set the NO_CACHE environment variable
to any non-zero value.
If you get errors about Wide characters in print it means that you've
forgotten to emit pure HTML on non-ASCII characters.
See L<HTML::Entities>.
As a hack work around you could also remove accents and the like by using
L<Text::Unidecode>,
which works well but isn't really what you want.

=head1 SUBROUTINES/METHODS

=cut

use constant MIN_GZIP_LEN => 32;

our $generate_etag = 1;
our $generate_304 = 1;
our $generate_last_modified = 1;
our $compress_content = 1;
our $optimise_content = 0;
our $lint_content = 0;
our $cache;
our $cache_age;
our $cache_key;
our $info;
our $logger;
our $lingua;
our $status;
our $script_mtime;
our $cobject;
our($x_cache, $buf, $headers, $header, $body, @content_type, $etag,
	$send_body, @o, $encode_loaded);

BEGIN {
	# use Exporter();

	$CGI::Buffer::buf = IO::String->new();
	$CGI::Buffer::old_buf = select($CGI::Buffer::buf);

	if((!defined($ENV{'SERVER_PROTOCOL'})) ||
	  ($ENV{'SERVER_PROTOCOL'} eq 'HTTP/1.0')) {
		$generate_etag = 0;
	}
}

END {
	if(defined($^V) && ($^V ge 'v5.14.0')) {
		return if ${^GLOBAL_PHASE} eq 'DESTRUCT';	# >= 5.14.0 only
	}

	if($logger) {
		if($ENV{'HTTP_IF_NONE_MATCH'}) {
			$logger->debug("HTTP_IF_NONE_MATCH: $ENV{HTTP_IF_NONE_MATCH}");
		}
		if($ENV{'HTTP_IF_MODIFIED_SINCE'}) {
			$logger->debug("HTTP_IF_MODIFIED_SINCE: $ENV{HTTP_IF_MODIFIED_SINCE}");
		}
		$logger->debug("Generate_etag = $generate_etag, ",
			"Generate_304 = $generate_304, ",
			"Generate_last_modified = $generate_last_modified");

		# This will cause everything to get flushed and prevent
		# outputs to the logger.  We need to do that now since
		# if we leave it to Perl to delete later we may get
		# a message that Log4perl::init() hasn't been called
		$logger = undef;
	}
	select($CGI::Buffer::old_buf);
	my $pos = $CGI::Buffer::buf->getpos;
	$CGI::Buffer::buf->setpos(0);
	read($CGI::Buffer::buf, $buf, $pos);
	($headers, $body) = split /\r?\n\r?\n/, $buf, 2;

	unless($headers || is_cached()) {
		if($logger) {
			$logger->debug('There was no output');
		}
		if(!defined($headers)) {
			require HTTP::Status;
			HTTP::Status->import();

			if(!defined($status)) {
				$status = 200;
			}
			print "Status: $status ",
				HTTP::Status::status_message($status),
				"\n\n";
		}
		return;
	}
	if($ENV{'REQUEST_METHOD'} && ($ENV{'REQUEST_METHOD'} eq 'HEAD')) {
		$send_body = 0;
	} else {
		$send_body = 1;
	}

	if($headers) {
		_set_content_type($headers);
	}

	if(defined($body) && ($body eq '')) {
		# E.g. if header of Location is given with no body, for
		#	redirection
		$body = undef;
		if($cache) {
			# Don't try to retrieve it below from the cache
			$send_body = 0;
		}
	} elsif(defined($content_type[0]) && (lc($content_type[0]) eq 'text') && (lc($content_type[1]) =~ /^html/) && defined($body)) {
		if($optimise_content) {
			# require HTML::Clean;
			require HTML::Packer;	# Overkill using HTML::Clean and HTML::Packer...

			if($logger) {
				$logger->trace('Packer');
			}

			my $oldlength = length($body);
			my $newlength;

			if($optimise_content == 1) {
				_optimise_content();
			} else {
				while(1) {
					_optimise_content();
					$newlength = length($body);
					last if ($newlength >= $oldlength);
					$oldlength = $newlength;
				}
			}

			# If we're on http://www.example.com and have a link
			# to http://www.example.com/foo/bar.htm, change the
			# link to /foo/bar.htm - there's no need to include
			# the site name in the link
			unless(defined($info)) {
				if($cache) {
					$info = CGI::Info->new({ cache => $cache });
				} else {
					$info = CGI::Info->new();
				}
			}

			my $href = $info->host_name();
			my $protocol = $info->protocol();

			unless($protocol) {
				$protocol = 'http';
			}

			$body =~ s/<a\s+?href="$protocol:\/\/$href"/<a href="\/"/gim;
			$body =~ s/<a\s+?href="$protocol:\/\/$href/<a href="/gim;

			# TODO use URI->path_segments to change links in
			# /aa/bb/cc/dd.htm which point to /aa/bb/ff.htm to
			# ../ff.htm

			# TODO: <img border=0 src=...>
			$body =~ s/<img\s+?src="$protocol:\/\/$href"/<img src="\/"/gim;
			$body =~ s/<img\s+?src="$protocol:\/\/$href/<img src="/gim;

			# Don't use HTML::Clean because of RT402
			# my $h = new HTML::Clean(\$body);
			# # $h->compat();
			# $h->strip();
			# my $ref = $h->data();

			# Don't always do javascript 'best' since it's confused
			# by the common <!-- HIDE technique.
			# See https://github.com/nevesenin/javascript-packer-perl/issues/1#issuecomment-4356790
			my $options = {
				remove_comments => 1,
				remove_newlines => 0,
				do_stylesheet => 'minify'
			};
			if($optimise_content >= 2) {
				$options->{do_javascript} = 'best';
				$body =~ s/(<script.*?>)\s*<!--/$1/gi;
				$body =~ s/\/\/-->\s*<\/script>/<\/script>/gi;
				$body =~ s/(<script.*?>)\s+/$1/gi;
			}
			$body = HTML::Packer->init()->minify(\$body, $options);
			if($optimise_content >= 2) {
				# Change document.write("a"); document.write("b")
				# into document.write("a"+"b");
				while(1) {
					$body =~ s/<script\s*?type\s*?=\s*?"text\/javascript"\s*?>(.*?)document\.write\((.+?)\);\s*?document\.write\((.+?)\)/<script type="text\/JavaScript">${1}document.write($2+$3)/igs;
					$newlength = length($body);
					last if ($newlength >= $oldlength);
					$oldlength = $newlength;
				}
			}
		}
		if($lint_content) {
			require HTML::Lint;
			HTML::Lint->import;

			if($logger) {
				$logger->trace('Lint');
			}
			my $lint = HTML::Lint->new();
			$lint->parse($body);
			$lint->eof();

			if($lint->errors) {
				$headers = 'Status: 500 Internal Server Error';
				@o = ('Content-type: text/plain');
				$body = '';
				foreach my $error ($lint->errors) {
					my $errtext = $error->where() . ': ' . $error->errtext() . "\n";
					if($logger) {
						$logger->warn($errtext);
					} else {
						warn($errtext);
					}
					$body .= $errtext;
				}
			}
		}
	}

	if(defined($headers) && ($headers =~ /^Status: (\d+)/m)) {
		$status = $1;
	} elsif($info) {
		$status = $info->status();
	} else {
		$status = 200;
	}

	if($logger) {
		$logger->debug("Initial status = $status");
	}

	# Generate the eTag before compressing, since the compressed data
	# includes the mtime field which changes thus causing a different
	# Etag to be generated
	if($ENV{'SERVER_PROTOCOL'} &&
	  (($ENV{'SERVER_PROTOCOL'} eq 'HTTP/1.1') || ($ENV{'SERVER_PROTOCOL'} eq 'HTTP/2.0')) &&
	  $generate_etag && defined($body)) {
		# encode to avoid "Wide character in subroutine entry"
		require Encode;
		$encode_loaded = 1;
		$etag = '"' . Digest::MD5->new->add(Encode::encode_utf8($body))->hexdigest() . '"';
		if($ENV{'HTTP_IF_NONE_MATCH'} && $generate_304 && ($status == 200)) {
			if($logger) {
				$logger->debug("Compare $ENV{HTTP_IF_NONE_MATCH} with $etag");
			}
			if($ENV{'HTTP_IF_NONE_MATCH'} eq $etag) {
				push @o, "Status: 304 Not Modified";
				$send_body = 0;
				$status = 304;
				if($logger) {
					$logger->debug('Set status to 304');
				}
			} elsif($logger) {
				$logger->debug(diff(\$body, \$cache->get(_generate_key())));
			}
		}
	}

	my $encoding = _should_gzip();
	my $unzipped_body = $body;

	if(defined($unzipped_body)) {
		my $range = $ENV{'Range'} ? $ENV{'Range'} : $ENV{'HTTP_RANGE'};

		if($range && !$cache) {
			# TODO: Partials
			if($range =~ /^bytes=(\d*)-(\d*)/) {
				if($1 && $2) {
					$body = substr($body, $1, $2-$1);
				} elsif($1) {
					$body = substr($body, $1);
				} elsif($2) {
					$body = substr($body, 0, $2);
				}
				$unzipped_body = $body;
				$status = 206;
			}
		}
		_compress({ encoding => $encoding });
	}

	if($cache) {
		require Storable;

		my $cache_hash;
		my $key = _generate_key();

		# Cache unzipped version
		if(!defined($body)) {
			if($send_body) {
				$cobject = $cache->get_object($key);
				if(defined($cobject)) {
					$cache_hash = Storable::thaw($cobject->value());
					$headers = $cache_hash->{'headers'};
					_set_content_type($headers);
					@o = ("X-CGI-Buffer-$VERSION: Hit");
					if($info) {
						my $host_name = $info->host_name();
						push @o, "X-Cache: HIT from $host_name";
						push @o, "X-Cache-Lookup: HIT from $host_name";
					} else {
						push @o, 'X-Cache: HIT';
						push @o, 'X-Cache-Lookup: HIT';
					}
				} elsif($logger) {
					$logger->warn("Error retrieving data for key $key");
				} else {
					carp(__PACKAGE__, ": error retrieving data for key $key");
				}
			}

			# Nothing has been output yet, so we can check if it's
			# OK to send 304 if possible
			if($send_body && $ENV{'SERVER_PROTOCOL'} &&
			  (($ENV{'SERVER_PROTOCOL'} eq 'HTTP/1.1') || ($ENV{'SERVER_PROTOCOL'} eq 'HTTP/2.0')) &&
			  $generate_304 && ($status == 200)) {
				if($ENV{'HTTP_IF_MODIFIED_SINCE'}) {
					_check_modified_since({
						since => $ENV{'HTTP_IF_MODIFIED_SINCE'},
						modified => $cobject->created_at()
					});
				}
			}
			if($send_body && ($status == 200) && defined($cache_hash)) {
				$body = $cache_hash->{'body'};
				if(!defined($body)) {
					# Panic
					$headers = 'Status: 500 Internal Server Error';
					@o = ('Content-type: text/plain');
					$body = "Can't retrieve body for key $key, cache_hash contains:\n";
					foreach my $k (keys %{$cache_hash}) {
						$body .= "\t$k\n";
					}
					$cache->remove($key);
					if($logger) {
						$logger->error("Can't retrieve body for key $key");
						$logger->warn($body);
					} else {
						carp "Can't retrieve body for key $key";
						warn($body);
					}
					$send_body = 0;
					$status = 500;
				}
			}
			if($send_body && $ENV{'SERVER_PROTOCOL'} &&
			  (($ENV{'SERVER_PROTOCOL'} eq 'HTTP/1.1') || ($ENV{'SERVER_PROTOCOL'} eq 'HTTP/2.0')) &&
			  ($status == 200)) {
				if($ENV{'HTTP_IF_NONE_MATCH'}) {
					if(!defined($etag)) {
						unless($encode_loaded) {
							require Encode;
							$encode_loaded = 1;
						}
						$etag = '"' . Digest::MD5->new->add(Encode::encode_utf8($body))->hexdigest() . '"';
					}
					if($logger && $generate_304) {
						$logger->debug("Compare etags $ENV{HTTP_IF_NONE_MATCH} and $etag");
					}
					if(($ENV{'HTTP_IF_NONE_MATCH'} eq $etag) && $generate_304) {
						push @o, "Status: 304 Not Modified";
						$status = 304;
						$send_body = 0;
						if($logger) {
							$logger->debug('Set status to 304');
						}
					}
				}
			}
			if($status == 200) {
				$encoding = _should_gzip();
				if($send_body) {
					if($generate_etag && !defined($etag) && defined($body) && ((!defined($headers)) || ($headers !~ /^ETag: /m))) {
						$etag = '"' . Digest::MD5->new->add(Encode::encode_utf8($body))->hexdigest() . '"';
					}
					_compress({ encoding => $encoding });
				}
			}
			my $cannot_304 = !$generate_304;
			unless($etag) {
				if(defined($headers) && ($headers =~ /^ETag: "([a-z0-9]{32})"/m)) {
					$etag = $1;
				} else {
					$etag = $cache_hash->{'etag'};
				}
			}
			if($ENV{'HTTP_IF_NONE_MATCH'} && $send_body && ($status != 304) && $generate_304) {
				if($logger) {
					$logger->debug("Compare $ENV{HTTP_IF_NONE_MATCH} with $etag");
				}
				if(defined($etag) && ($etag eq $ENV{'HTTP_IF_NONE_MATCH'}) && ($status == 200)) {
					push @o, "Status: 304 Not Modified";
					$send_body = 0;
					$status = 304;
					if($logger) {
						$logger->debug('Set status to 304');
					}
				} else {
					$cannot_304 = 1;
				}
			}
			if($cobject) {
				if($ENV{'HTTP_IF_MODIFIED_SINCE'} && ($status != 304) && !$cannot_304) {
					_check_modified_since({
						since => $ENV{'HTTP_IF_MODIFIED_SINCE'},
						modified => $cobject->created_at()
					});
				}
				if(($status == 200) && $generate_last_modified) {
					if($logger) {
						$logger->debug('Set Last-Modified to ', HTTP::Date::time2str($cobject->created_at()));
					}
					push @o, "Last-Modified: " . HTTP::Date::time2str($cobject->created_at());
				}
			}
		} else {
			# Not in the server side cache
			if($status == 200) {
				unless($cache_age) {
					# It would be great if CHI::set()
					# allowed the time to be 'lru' for least
					# recently used.
					$cache_age = '10 minutes';
				}
				$cache_hash->{'body'} = $unzipped_body;
				if(@o && defined($o[0])) {
					# Remember, we're storing the UNzipped
					# version in the cache
					my $c;
					if(defined($headers) && length($headers)) {
						$c = $headers . "\r\n" . join("\r\n", @o);
					} else {
						$c = join("\r\n", @o);
					}
					$c =~ s/^Content-Encoding: .+$//mg;
					$c =~ s/^Vary: Accept-Encoding.*\r?$//mg;
					$c =~ s/\n+/\n/gs;
					if(length($c)) {
						$cache_hash->{'headers'} = $c;
					}
				} elsif(defined($headers) && length($headers)) {
					$headers =~ s/^Content-Encoding: .+$//mg;
					$headers =~ s/^Vary: Accept-Encoding.*\r?$//mg;
					$headers =~ s/\n+/\n/gs;
					if(length($headers)) {
						$cache_hash->{'headers'} = $headers;
					}
				}
				if($generate_etag && defined($etag)) {
					$cache_hash->{'etag'} = $etag;
				}
				# TODO: Support the Expires header
				# if($headers !~ /^Expires: /m))) {
				# }
				if($logger) {
					$logger->debug("Store $key in the cache, age = $cache_age ", length($cache_hash->{'body'}), ' bytes');
				}
				$cache->set($key, Storable::freeze($cache_hash), $cache_age);
				if($generate_last_modified) {
					$cobject = $cache->get_object($key);
					if(defined($cobject)) {
						push @o, "Last-Modified: " . HTTP::Date::time2str($cobject->created_at());
					} else {
						push @o, "Last-Modified: " . HTTP::Date::time2str(time);
					}
				}
			}
			if($info) {
				my $host_name = $info->host_name();
				if(defined($x_cache)) {
					push @o, "X-Cache: $x_cache from $host_name";
				} else {
					push @o, "X-Cache: MISS from $host_name";
				}
				push @o, "X-Cache-Lookup: MISS from $host_name";
			} else {
				if(defined($x_cache)) {
					push @o, "X-Cache: $x_cache";
				} else {
					push @o, 'X-Cache: MISS';
				}
				push @o, 'X-Cache-Lookup: MISS';
			}
			push @o, "X-CGI-Buffer-$VERSION: Miss";
		}
		# We don't need it any more, so give Perl a chance to
		# tidy it up seeing as we're in the destructor
		$cache = undef;
	} elsif($info) {
		my $host_name = $info->host_name();
		push @o, ("X-Cache: MISS from $host_name", "X-Cache-Lookup: MISS from $host_name");
		if($generate_last_modified) {
			if(my $age = _my_age()) {
				push @o, 'Last-Modified: ' . HTTP::Date::time2str($age);
			}
		}
		if($ENV{'HTTP_IF_MODIFIED_SINCE'} && ($status != 304) && $generate_304) {
			_check_modified_since({
				since => $ENV{'HTTP_IF_MODIFIED_SINCE'},
				modified => _my_age()
			});
		}
	} else {
		push @o, ('X-Cache: MISS', 'X-Cache-Lookup: MISS');
	}
	if($generate_etag && ((!defined($headers)) || ($headers !~ /^ETag: /m))) {
		if(defined($etag)) {
			push @o, "ETag: $etag";
			if($logger) {
				$logger->debug("Set ETag to $etag");
			}
		} elsif($logger && (($status == 200) || $status == 304) && $body && !is_cached()) {
			$logger->warn("BUG: ETag not generated, status $status");
		}
	}

	my $body_length;
	if(defined($body)) {
		if(utf8::is_utf8($body)) {
			utf8::encode($body);
		}
		$body_length = length($body);
	} else {
		$body_length = 0;
	}

	if(defined($headers) && length($headers)) {
		# Put the original headers first, then those generated within
		# CGI::Buffer
		unshift @o, split(/\r\n/, $headers);
		if($body && $send_body) {
			if(scalar(grep(/^Content-Length: \d/, @o)) == 0) {
				push @o, "Content-Length: $body_length";
			}
		}
		if(scalar(grep(/^Status: \d/, @o)) == 0) {
			require HTTP::Status;
			HTTP::Status->import();

			push @o, "Status: $status " . HTTP::Status::status_message($status);
		}
	} else {
		push @o, "X-CGI-Buffer-$VERSION: No headers";
	}

	if($body_length && $send_body) {
		push @o, ('', $body);
	}

	# XXXXXXXXXXXXXXXXXXXXXXX
	if(0) {
		# This code helps to debug Wide character prints
		my $wideCharWarningsIssued = 0;
		my $widemess;
		$SIG{__WARN__} = sub {
			$wideCharWarningsIssued += "@_" =~ /Wide character in .../;
			$widemess = "@_";
			if($logger) {
				$logger->fatal($widemess);
				my $i = 1;
				$logger->trace('Stack Trace');
				while((my @call_details = (caller($i++)))) {
					$logger->trace($call_details[1] . ':' . $call_details[2] . ' in function ' . $call_details[3]);
				}
			}
			CORE::warn(@_);	# call the builtin warn as usual
		};

		if(scalar @o) {
			print join("\r\n", @o);
			if($wideCharWarningsIssued) {
				my $mess = join("\r\n", @o);
				$mess =~ /[^\x00-\xFF]/;
				open(my $fout, '>>', '/tmp/NJH');
				print $fout "$widemess:\n";
				print $fout $mess;
				print $fout 'x' x 40, "\n";
				close $fout;
			}
		}
	} elsif(scalar @o) {
		print join("\r\n", @o);
	}
	# XXXXXXXXXXXXXXXXXXXXXXX

	if((!$send_body) || !defined($body)) {
		print "\r\n\r\n";
	}
}

sub _check_modified_since {
	if($logger) {
		$logger->trace('In _check_modified_since');
	}

	if(!$generate_304) {
		return;
	}
	my $params = shift;

	if(!defined($$params{since})) {
		return;
	}
	my $s = HTTP::Date::str2time($$params{since});
	if(!defined($s)) {
		# IF_MODIFIED_SINCE isn't a valid data
		return;
	}

	my $age = _my_age();
	if(!defined($age)) {
		return;
	}
	if($age > $s) {
		if($logger) {
			$logger->debug('_check_modified_since: script has been modified');
		}
		# Script has been updated so it may produce different output
		return;
	}

	if($logger) {
		$logger->debug("_check_modified_since: Compare $$params{modified} with $s");
	}
	if($$params{modified} <= $s) {
		push @o, "Status: 304 Not Modified";
		$status = 304;
		$send_body = 0;
		if($logger) {
			$logger->debug('Set status to 304');
		}
	}
}

# Reduce output, e.g. remove superfluous white-space.
sub _optimise_content {
	# FIXME: regex bad, HTML parser good
	# Regexp::List - wow!
	$body =~ s/(((\s+|\r)\n|\n(\s+|\+)))/\n/g;
	# $body =~ s/\r\n/\n/gs;
	# $body =~ s/\s+\n/\n/gs;
	# $body =~ s/\n+/\n/gs;
	# $body =~ s/\n\s+|\s+\n/\n/g;
	$body =~ s/\<\/div\>\s+\<div/\<\/div\>\<div/gis;
	# $body =~ s/\<\/p\>\s\<\/div/\<\/p\>\<\/div/gis;
	# $body =~ s/\<div\>\s+/\<div\>/gis;	# Remove spaces after <div>
	$body =~ s/(<div>\s+|\s+<div>)/<div>/gis;
	$body =~ s/\s+<\/div\>/\<\/div\>/gis;	# Remove spaces before </div>
	$body =~ s/\s+\<p\>|\<p\>\s+/\<p\>/im;	# TODO <p class=
	$body =~ s/\s+\<\/p\>|\<\/p\>\s+/\<\/p\>/gis;
	$body =~ s/<html>\s+<head>/<html><head>/is;
	$body =~ s/\s*<\/head>\s+<body>\s*/<\/head><body>/is;
	$body =~ s/<html>\s+<body>/<html><body>/is;
	$body =~ s/<body>\s+/<body>/is;
	$body =~ s/\s+\<\/html/\<\/html/is;
	$body =~ s/\s+\<\/body/\<\/body/is;
	$body =~ s/\s(\<.+?\>\s\<.+?\>)/$1/;
	# $body =~ s/(\<.+?\>\s\<.+?\>)\s/$1/g;
	$body =~ s/\<p\>\s/\<p\>/gi;
	$body =~ s/\<\/p\>\s\<p\>/\<\/p\>\<p\>/gi;
	$body =~ s/\<\/tr\>\s\<tr\>/\<\/tr\>\<tr\>/gi;
	$body =~ s/\<\/td\>\s\<\/tr\>/\<\/td\>\<\/tr\>/gi;
	$body =~ s/\<\/td\>\s*\<td\>/\<\/td\>\<td\>/gis;
	$body =~ s/\<\/tr\>\s\<\/table\>/\<\/tr\>\<\/table\>/gi;
	$body =~ s/\<br\s?\/?\>\s?\<p\>/\<p\>/gi;
	$body =~ s/\<br\>\s/\<br\>/gi;
	$body =~ s/\s+\<br/\<br/gi;
	$body =~ s/\<br\s?\/\>\s/\<br \/\>/gi;
	$body =~ s/[ \t]+/ /gs;	# Remove duplicate space, don't use \s+ it breaks JavaScript
	$body =~ s/\s\<p\>/\<p\>/gi;
	$body =~ s/\s\<script/\<script/gi;
	$body =~ s/(<script>\s|\s<script>)/<script>/gis;
	$body =~ s/(<\/script>\s|\s<\/script>)/<\/script>/gis;
	$body =~ s/\<td\>\s/\<td\>/gi;
	$body =~ s/\s+\<a\shref="(.+?)"\>\s?/ <a href="$1">/gis;
	$body =~ s/\s?<a\shref=\s"(.+?)"\>/ <a href="$1">/gis;
	$body =~ s/\s+<\/a\>\s+/<\/a> /gis;
	$body =~ s/(\s?<hr>\s+|\s+<hr>\s?)/<hr>/gis;
	# $body =~ s/\s<hr>/<hr>/gis;
	# $body =~ s/<hr>\s/<hr>/gis;
	$body =~ s/<\/li>\s+<li>/<\/li><li>/gis;
	$body =~ s/<\/li>\s+<\/ul>/<\/li><\/ul>/gis;
	$body =~ s/<ul>\s+<li>/<ul><li>/gis;
	$body =~ s/\s+<\/li>/<\/li>/gis;
	$body =~ s/\<\/option\>\s+\<option/\<\/option\>\<option/gis;
	$body =~ s/<title>\s*(.+?)\s*<\/title>/<title>$1<\/title>/is;
	$body =~ s/<\/center>\s+<center>/ /gis;
}

# Create a key for the cache
sub _generate_key {
	if($cache_key) {
		return $cache_key;
	}
	unless(defined($info)) {
		$info = CGI::Info->new({ cache => $cache });
	}

	my $key = $info->browser_type() . '::' . $info->domain_name() . '::' . $info->script_name() . '::' . $info->as_string();
	if($lingua) {
		$key .= '::' . $lingua->language();
	}
	if($ENV{'HTTP_COOKIE'}) {
		# Different states of the client are stored in different caches
		# Don't put different Google Analytics in different caches, and anyway they
		# would be wrong
		foreach my $cookie(split(/;/, $ENV{'HTTP_COOKIE'})) {
			unless($cookie =~ /^__utm[abcz]/) {
				$key .= "::$cookie";
			}
		}
	}

	# Honour the Vary headers
	if($headers && ($headers =~ /^Vary: .*$/m)) {
		if(defined($logger)) {
			$logger->debug('Found Vary header');
		}
		foreach my $h1(split(/\r?\n/, $headers)) {
			my ($h1_name, $h1_value) = split /\:\s*/, $h1, 2;
			if(lc($h1_name) eq 'vary') {
				foreach my $h2(split(/\r?\n/, $headers)) {
					my ($h2_name, $h2_value) = split /\:\s*/, $h2, 2;
					if($h2_name eq $h1_value) {
						$key .= '::' . $h2_value;
						last;
					}
				}
			}
		}
	}
	$key =~ s/\//::/g;
	$key =~ s/::::/::/g;
	$key =~ s/::$//;
	if(defined($logger)) {
		$logger->trace("Returning $key");
	}
	$cache_key = $key;
	return $key;
}

=head2 init

Set various options and override default values.

    # Put this toward the top of your program before you do anything
    # By default, generate_tag, generate_304 and compress_content are ON,
    # optimise_content and lint_content are OFF.  Set optimise_content to 2 to
    # do aggressive JavaScript optimisations which may fail.
    use CGI::Buffer;
    CGI::Buffer::init(
	generate_etag => 1,	# make good use of client's cache
	generate_last_modified => 1,	# more use of client's cache
	compress_content => 1,	# if gzip the output
	optimise_content => 0,	# optimise your program's HTML, CSS and JavaScript
	cache => CHI->new(driver => 'File'),	# cache requests
	cache_key => 'string',	# key for the cache
	cache_age => '10 minutes',	# how long to store responses in the cache
	logger => $logger,
	lint_content => 0,	# Pass through HTML::Lint
	generate_304 => 1,	# Generate 304: Not modified
	lingua => CGI::Lingua->new(),
    );

If no cache_key is given, one will be generated which may not be unique.
The cache_key should be a unique value dependent upon the values set by the
browser.

The cache object will be an object that understands get_object(),
set(), remove() and created_at() messages, such as an L<CHI> object. It is
used as a server-side cache to reduce the need to rerun database accesses.

Items stay in the server-side cache by default for 10 minutes.
This can be overridden by the cache_control HTTP header in the request, and
the default can be changed by the cache_age argument to init().

Logger will be an object that understands debug() such as an L<Log::Log4perl>
object.

To generate a last_modified header, you must give a cache object.

Init allows a reference of the options to be passed. So both of these work:
    use CGI::Buffer;
    #...
    CGI::Buffer::init(generate_etag => 1);
    CGI::Buffer::init({ generate_etag => 1, info => CGI::Info->new() });

Generally speaking, passing by reference is better since it copies less on to
the stack.

Alternatively you can give the options when loading the package:
    use CGI::Buffer { optimise_content => 1 };

=cut

sub init {
	my %params = (ref($_[0]) eq 'HASH') ? %{$_[0]} : @_;

	# Safe options - can be called at any time
	if(defined($params{generate_etag})) {
		$generate_etag = $params{generate_etag};
	}
	if(defined($params{generate_last_modified})) {
		$generate_last_modified = $params{generate_last_modified};
	}
	if(defined($params{compress_content})) {
		$compress_content = $params{compress_content};
	}
	if(defined($params{optimise_content})) {
		$optimise_content = $params{optimise_content};
	}
	if(defined($params{lint_content})) {
		$lint_content = $params{lint_content};
	}
	if(defined($params{logger})) {
		$logger = $params{logger};
	}
	if(defined($params{lingua})) {
		$lingua = $params{lingua};
	}
	if(defined($params{generate_304})) {
		$generate_304 = $params{generate_304};
	}
	if(defined($params{info}) && (!defined($info))) {
		$info = $params{info};
	}

	# Unsafe options - must be called before output has been started
	my $pos = $CGI::Buffer::buf->getpos;
	if($pos > 0) {
		if(defined($logger)) {
			my @call_details = caller(0);
			$logger->warn("Too late to call init, $pos characters have been printed, caller line $call_details[2] of $call_details[1]");
		} else {
			# Must do Carp::carp instead of carp for Test::Carp
			Carp::carp "Too late to call init, $pos characters have been printed";
		}
	}
	if(defined($params{cache}) && can_cache()) {
		if(defined($ENV{'HTTP_CACHE_CONTROL'})) {
			my $control = $ENV{'HTTP_CACHE_CONTROL'};
			if(defined($logger)) {
				$logger->debug("cache_control = $control");
			}
			if($control =~ /^max-age\s*=\s*(\d+)$/) {
				# There is an argument not to do this
				# since one client will affect others
				$cache_age = "$1 seconds";
				if(defined($logger)) {
					$logger->debug("cache_age = $cache_age");
				}
			}
		}
		$cache_age ||= $params{cache_age};

		if((!defined($params{cache})) && defined($cache)) {
			if(defined($logger)) {
				if($cache_key) {
					$logger->debug("disabling cache $cache_key");
				} else {
					$logger->debug('disabling cache');
				}
			}
			$cache = undef;
		} else {
			$cache = $params{cache};
		}
		if(defined($params{cache_key})) {
			$cache_key = $params{cache_key};
		}
	}
}

sub import {
	# my $class = shift;
	shift;

	return unless @_;

	init(@_);
}

=head2 set_options

Synonym for init, kept for historical reasons.

=cut

sub set_options {
	my %params = (ref($_[0]) eq 'HASH') ? %{$_[0]} : @_;

	init(\%params);
}

=head2 can_cache

Returns true if the server is allowed to store the results locally.

=cut

sub can_cache {
	if(defined($x_cache)) {
		return ($x_cache eq 'HIT');
	}

	if(defined($ENV{'NO_CACHE'}) || defined($ENV{'NO_STORE'})) {
		$x_cache = 'MISS';
		return 0;
	}
	if(defined($ENV{'HTTP_CACHE_CONTROL'})) {
		my $control = $ENV{'HTTP_CACHE_CONTROL'};
		if(defined($logger)) {
			$logger->debug("cache_control = $control");
		}
		# TODO: check Authorization header not present
		if(($control eq 'no-store') ||
		       ($control eq 'no-cache') ||
		       ($control eq 'max-age=0') ||
		       ($control eq 'private')) {
			$x_cache = 'MISS';
			return 0;
		}
	}
	$x_cache = 'HIT';
	return 1;
}

=head2 is_cached

Returns true if the output is cached. If it is then it means that all of the
expensive routines in the CGI script can be by-passed because we already have
the result stored in the cache.

    # Put this toward the top of your program before you do anything

    # Example key generation - use whatever you want as something
    # unique for this call, so that subsequent calls with the same
    # values match something in the cache
    use CGI::Info;
    use CGI::Lingua;
    use CGI::Buffer;

    my $i = CGI::Info->new();
    my $l = CGI::Lingua->new(supported => ['en']);

    # To use server side caching you must give the cache argument, however
    # the cache_key argument is optional - if you don't give one then one will
    # be generated for you
    if(CGI::Buffer::can_cache()) {
        CGI::Buffer::init(
	    cache => CHI->new(driver => 'File'),
	    cache_key => $i->domain_name() . '/' . $i->script_name() . '/' . $i->as_string() . '/' . $l->language()
        );
        if(CGI::Buffer::is_cached()) {
	    # Output will be retrieved from the cache and sent automatically
	    exit;
        }
    }
    # Not in the cache, so now do our expensive computing to generate the
    # results
    print "Content-type: text/html\n";
    # ...

=cut

sub is_cached {
	unless($cache) {
		if($logger) {
			$logger->debug("is_cached: cache hasn't been enabled");
		}
		return 0;
	}

	my $key = _generate_key();

	if($logger) {
		$logger->debug("is_cached: looking for key = $key");
	}
	$cobject = $cache->get_object($key);
	unless($cobject) {
		if($logger) {
			$logger->debug('not found in cache');
		}
		return 0;
	}
	unless($cobject->value($key)) {
		if($logger) {
			$logger->warn('is_cached: object is in the cache but not the data');
		}
		$cobject = undef;
		return 0;
	}

	# If the script has changed, don't use the cache since we may produce
	# different output
	my $age = _my_age();
	unless(defined($age)) {
		if($logger) {
			$logger->debug("Can't determine script's age");
		}
		# Can't determine the age. Play it safe an assume we're not
		# cached
		$cobject = undef;
		return 0;
	}
	if($age > $cobject->created_at()) {
		# Script has been updated so it may produce different output
		if($logger) {
			$logger->debug('Script has been updated');
		}
		$cobject = undef;
		# Nothing will be in date and all new searches would miss
		# anyway, so may as well clear it all
		# FIXME: RT104471
		# $cache->clear();
		return 0;
	}
	if($logger) {
		$logger->debug('Script is in the cache');
	}
	return 1;
}

sub _my_age {
	if($script_mtime) {
		return $script_mtime;
	}
	unless(defined($info)) {
		if($cache) {
			$info = CGI::Info->new({ cache => $cache });
		} else {
			$info = CGI::Info->new();
		}
	}

	my $path = $info->script_path();
	unless(defined($path)) {
		return;
	}

	my @statb = stat($path);
	$script_mtime = $statb[9];
	return $script_mtime;
}

sub _should_gzip
{
	if($compress_content && ($ENV{'HTTP_ACCEPT_ENCODING'} || $ENV{'HTTP_TE'})) {
		if(scalar(@content_type)) {
			if($content_type[0] ne 'text') {
				return '';
			}
		}
		my $accept = lc($ENV{'HTTP_ACCEPT_ENCODING'} ? $ENV{'HTTP_ACCEPT_ENCODING'} : $ENV{'HTTP_TE'});
		foreach my $method(split(/,\s?/, $accept)) {
			if(($method eq 'gzip') || ($method eq 'x-gzip') || ($method eq 'br')) {
				return $method;
			}
		}
	}

	return '';
}

sub _set_content_type
{
	my $headers = shift;

	foreach my $header (split(/\r?\n/, $headers)) {
		my ($header_name, $header_value) = split /\:\s*/, $header, 2;
		if (lc($header_name) eq 'content-type') {
			@content_type = split /\//, $header_value, 2;
			last;
		}
	}
}

sub _compress {
	my %params = (ref($_[0]) eq 'HASH') ? %{$_[0]} : @_;

	return unless(defined($body));

	my $encoding = $params{encoding};

	if((length($encoding) == 0) || (length($body) < MIN_GZIP_LEN)) {
		return;
	}

	if($encoding eq 'gzip') {
		require Compress::Zlib;
		Compress::Zlib->import;

		# Avoid 'Wide character in memGzip'
		unless($encode_loaded) {
			require Encode;
			$encode_loaded = 1;
		}
		my $nbody = Compress::Zlib::memGzip(\Encode::encode_utf8($body));
		if(length($nbody) < length($body)) {
			$body = $nbody;
			push @o, "Content-Encoding: $encoding";
			push @o, "Vary: Accept-Encoding";
		}
	} elsif($encoding eq 'br') {
		require IO::Compress::Brotli;
		IO::Compress::Brotli->import();

		# Avoid 'Wide character in memGzip'
		unless($encode_loaded) {
			require Encode;
			$encode_loaded = 1;
		}
		my $nbody = IO::Compress::Brotli::bro(Encode::encode_utf8($body));
		if(length($nbody) < length($body)) {
			$body = $nbody;
			push @o, "Content-Encoding: $encoding";
			push @o, "Vary: Accept-Encoding";
		}
	}
}

=head1 AUTHOR

Nigel Horne, C<< <njh at bandsman.co.uk> >>

=head1 BUGS

CGI::Buffer should be safe even in scripts which produce lots of different
output, e.g. e-commerce situations.
On such pages, however, I strongly urge to setting generate_304 to 0 and
sending the HTTP header "Cache-Control: no-cache".

When using L<Template>, ensure that you don't use it to output to STDOUT,
instead you will need to capture into a variable and print that.
For example:

    my $output;
    $template->process($input, $vars, \$output) || ($output = $template->error());
    print $output;

Can produce buggy JavaScript if you use the <!-- HIDING technique.
This is a bug in L<JavaScript::Packer>, not CGI::Buffer.
See https://github.com/nevesenin/javascript-packer-perl/issues/1#issuecomment-4356790

Mod_deflate can confuse this when compressing output.
Ensure that deflation is off for .pl files:

    SetEnvIfNoCase Request_URI \.(?:gif|jpe?g|png|pl)$ no-gzip dont-vary

If you request compressed output then uncompressed output (or vice
versa) on input that produces the same output, the status will be 304.
The letter of the spec says that's wrong, so I'm noting it here, but
in practice you should not see this happen or have any difficulties
because of it.

CGI::Buffer is not compatible with FastCGI.

I advise adding CGI::Buffer as the last use statement so that it is
cleared up first.  In particular it should be loaded after
L<Log::Log4perl>, if you're using that, so that any messages it
produces are printed after the HTTP headers have been sent by
CGI::Buffer;

CGI::Buffer is not compatible with FCGI, use L<FCGI::Buffer> instead.

Please report any bugs or feature requests to C<bug-cgi-buffer at rt.cpan.org>,
or through the web interface at L<http://rt.cpan.org/NoAuth/ReportBug.html?Queue=CGI-Buffer>.
I will be notified, and then you'll automatically be notified of progress on
your bug as I make changes.

=head1 SEE ALSO

L<HTML::Packer>, L<HTML::Lint>

=head1 SUPPORT

You can find documentation for this module with the perldoc command.

    perldoc CGI::Buffer

You can also look for information at:

=over 4

=item * MetaCPAN

L<https://metacpan.org/release/CGI-Buffer>

=item * RT: CPAN's request tracker

L<https://rt.cpan.org/NoAuth/Bugs.html?Dist=CGI-Buffer>

=item * CPANTS

L<http://cpants.cpanauthors.org/dist/CGI-Buffer>

=item * CPAN Testers' Matrix

L<http://matrix.cpantesters.org/?dist=CGI-Buffer>

=item * CPAN Ratings

L<http://cpanratings.perl.org/d/CGI-Buffer>

=item * CPAN Testers Dependencies

L<http://deps.cpantesters.org/?module=CGI::Buffer>

=back

=head1 ACKNOWLEDGEMENTS

The inspiration and code for some of this is cgi_buffer by Mark
Nottingham: L<https://www.mnot.net/blog/2003/04/24/etags>.

=head1 LICENSE AND COPYRIGHT

The licence for cgi_buffer is:

    "(c) 2000 Copyright Mark Nottingham <[e-mail]>

    This software may be freely distributed, modified and used,
    provided that this copyright notice remain intact.

    This software is provided 'as is' without warranty of any kind."

The rest of the program is Copyright 2011-2023 Nigel Horne,
and is released under the following licence: GPL2

=cut

1; # End of CGI::Buffer